Increased Mortality Among Persons With Chronic Hepatitis C With Moderate or Severe Liver Disease: A Cohort Study

Summary

Individuals with mild/moderate liver disease due to chronic hepatitis C virus are often denied treatment, a practice our data did not support. In these individuals, we found increased mortality and inability to predict disease progression with sufficiently high diagnostic accuracy.

Abstract

Background.

Despite the availability of curative treatment for hepatitis C virus (HCV) infection, because of cost, treatment is often denied until liver fibrosis has progressed to at least moderate fibrosis and, in some cases, cirrhosis. That practice is justified on assumptions that there are no medical consequences to having moderate disease and that disease stage transitions can be anticipated.

Methods.

We performed transient elastography on 964 people chronically infected with HCV with a history of injection drug use living in Baltimore, Maryland. Liver stiffness was evaluated semiannually from 2006 to 2014 using validated cutoffs for moderate fibrosis (8.0–12.3 kPa) and severe fibrosis/cirrhosis (>12.3 kPa).

Results.

Among 964 persons, 62%, 23% and 15% had baseline measurements suggestive of no/mild fibrosis, moderate fibrosis, and severe fibrosis/cirrhosis, respectively. All-cause and nonaccidental mortality were elevated in persons with moderate fibrosis (adjusted hazard ratio [aHR], 1.42 [95% confidence interval {CI}, .96–2.11]; aHR, 1.66 [95% CI, 1.06–2.59], respectively) after adjustment for sociodemographics, substance use, and human immunodeficiency virus status. Despite the increased risk of mortality among those with moderate fibrosis, no combination of demographic, behavioral, and clinical factors, nor changes in stiffness measurements themselves could predict the transition from mild to moderate fibrosis with sufficiently high diagnostic accuracy (C-statistic = 0.72 for best-performing model).

Conclusions.

Delaying treatment for anyone chronically infected with HCV regardless of fibrosis stage may be detrimental given the increased risk of mortality even for those with moderate disease and the inability to predict the transition from mild to moderate disease.

Approximately 3 million people are chronically infected with hepatitis C virus (HCV) in the United States [1]; however, the true burden may be underestimated by at least several hundred thousand [2]. Chronic HCV infection is a major risk factor for liver cirrhosis, liver failure, and hepatocellular carcinoma [3]. Treatment with highly efficacious, oral direct-acting antivirals (DAAs) is curative [4]; however, barriers to access exist. As of December 2014, of 42 states that specified reimbursement procedures for DAAs, 31 restricted DAAs to patients with advanced fibrosis (METAVIR stage F3) or cirrhosis (METAVIR F4) [5], despite recommendations from the American Association for the Study of Liver Diseases (AASLD) and the Infectious Diseases Society of America (IDSA) to treat nearly everyone chronically infected [6]. Treatment denials are justified by the assumption of no medical consequence to low-stage HCV infection and that liver fibrosis progression can be safely monitored until advanced fibrosis/cirrhosis is detected.

Because guidelines recommend treatment for all, randomized trials to test the safety of treatment deferral for persons without advanced fibrosis/cirrhosis would be unethical. Thus, we characterize mortality rates by liver disease severity before widespread availability of DAAs in a community-based cohort of former/current people who inject drugs (PWID) in whom fibrosis was uniformly monitored with semiannual visits using transient elastography. We further examined whether liver disease progression from no/mild to moderate fibrosis could be predicted with high diagnostic accuracy. This is an aging cohort whose experience might be relevant to a large subgroup of persons in the United States who acquired HCV from drug use and were born between 1945 and 1965 [7, 8].

METHODS

Study Population

Data came from participants enrolled in the ALIVE (AIDS Linked to the IntraVenous Experience) study, a prospective community-recruited cohort of former and current PWID residing in Baltimore, Maryland [9]. Enrollment began in 1988, with additional recruitment during 1994–1995, 1998, 2000, and 2005–2008. Participants visited the clinic biannually, were interviewed, and provided biospecimens. All participants were ≥18 years old and had a history of injecting drugs. We limited this analysis to those with detectable HCV RNA who had at least 1 liver stiffness measurement (LSM). During the time period of the mortality analysis (until 2012), a negligible proportion of the cohort (<5%) reported receiving curative HCV treatment. The Johns Hopkins University institutional review board approved the study and all participants provided informed written consent.

Measurements

Liver Stiffness Measurements

Determination of liver fibrosis and cirrhosis was based on semiannual LSM by transient elastography using a FibroScan machine (Echosens, Paris, France), from 2006 until 2014. In brief, a transducer on the end of an ultrasound probe emits a shear wave and instantaneously records its velocity in kilopascals (kPa) as it passes through the liver. Trained personnel operated the machine in a research clinic. Valid examinations consisted of ≥8 valid measurements with a success rate of ≥60% (number of valid divided by total measurements) and limited variability (interquartile range [IQR] divided by the median <0.30) [8, 10]. We used the median from each valid examination. No/mild fibrosis, moderate fibrosis, and severe fibrosis/cirrhosis were defined by an LSM of <8.0 (METAVIR F0–F1), 8.0–12.3 (METAVIR F2–F3), and >12.3 kPa (METAVIR F4), respectively, according to validated cutoffs [10, 11].

Risk Behaviors and Serological Testing

Data on risk behaviors were collected at baseline and at 6-month follow-up visits. Comorbidities were assessed by either self-report of physician diagnosis or evidence of hypertension defined by a systolic blood pressure >150 mm Hg and diastolic blood pressure >90 mm Hg; or diabetes if hemoglobin A1C level was >6.5% [12]. The Alcohol Use Disorder Identification Test (AUDIT) was used to assess hazardous drinking/dependence [13].

Routine human immunodeficiency virus (HIV) serologic testing was conducted for participants not known to have HIV. Antibodies to HCV and hepatitis B virus surface antigen were obtained at the first available visit. The Abbott RealTime HCV assay (Abbott Molecular, Des Plaines, Illinois) assessed HCV RNA level at the visit approximately when the first LSM was performed. Platelet count, alanine aminotransferase, and aspartate aminotransferase were collected and used to calculate the Fibrosis-4 score (FIB-4) [14] within 1 year of the baseline LSM.

Statistical Analyses

Mortality

All HCV antibody–positive persons with detectable HCV RNA and at least 1 LSM were included (N = 964). Baseline for this analysis was the date of the first valid LSM. Mortality data, including data on cause, were obtained from the National Death Index (NDI) with confirmation from death certificates. Kaplan-Meier curves and Cox proportional hazards regression were used to compare mortality by baseline LSM. Persons were censored at their date of death or December 2012 (through which NDI data were complete). Selection of potential confounding variables included in multivariable regression models was based on previous mortality analyses in this cohort [12]. Two outcomes were considered: (1) all-cause mortality, and (2) nonaccidental mortality where deaths due to trauma or chronic drug and alcohol use were censored as nonevents. For both outcomes, we created 4 models: model 1, sociodemographics; model 2, sociodemographics and substance use; model 3, sociodemographics, substance use, and HIV infection; model 4, sociodemographics, substance use, HIV infection, and comorbidities (diabetes, hypertension). We assessed for violation of the proportional hazards assumption by examining interactions between covariates and time (none detected).

Predictors of Disease Progression

To characterize liver disease progression and identify whether disease progression could be predicted by risk factors or visit-to-visit changes in LSM, we restricted analyses to those with no/mild fibrosis (eg, LSM <8.0 kPa at first 2 visits [n = 331]), and thus not eligible for treatment in most settings. We used Cox regression to characterize associations between previously defined risk factors for progression [8]. An event was defined as ≥2 LSMs consistent with moderate fibrosis (≥8.0 kPa) [10]. Individuals who never had 2 LSMs ≥8.0 kPa during follow-up were censored. Model building was conducted using the same framework as the mortality analysis; however, behavioral factors and body mass index (BMI) were included as time-varying. We included visit-to-visit rates of change in LSM (∆) using 3 designations to represent typical times between clinical evaluations: T₀: first LSM ≥8.0 kPa (event); T_-1: 1 visit (approximately 6 months) prior to first LSM ≥8.0, T_-2: 2 visits (approximately 12 months) prior to first LSM ≥8.0, T_-3: 3 visits (approximately 18 months) prior to first LSM ≥8.0 (Figure 1) and all combinations in changes between visits (∆T_-1, T_-2; ∆T_-2, T_-3; ∆T_-1, T_-3). Change was calculated as the difference in the 2 LSM divided by the time between measurements. Thus, the analysis of visit-to-visit changes was limited to individuals with at least 4 LSMs (N = 331, 2719 LSMs). Three separate models were run for each distinct visit-to-visit change to assess whether an individual visit pair was associated with future fibrosis progression. All models were adjusted for the LSM value T_-3 (starting value). Most visits were between 6 and 12 months apart (median, 6.2 months [IQR, 6.0–11.9 months]) and the median change in LSM between visits was 0 kPa (IQR, –1.0 to 1.0 kPa). No significant difference in visit-to-visit LSM change was detected between visits that occurred every 6–12 months and visits that occurred >12 months apart (P = .41). Given potential differences on fibrosis progression due to HIV infection, we conducted sensitivity analyses stratifying by HIV status (Supplementary Material).

Figure 1.

Schematic of liver stiffness measurement (LSM) changes in between visits. *At least 2 visits ≥ 8.0 kPa.

Harrell C-statistic was calculated to assess the models’ ability to discriminate between those with and without evidence of disease progression, while allowing for inclusion of time- varying covariates and censored data [15, 16]. Methods described by Chambless et al [17] were used to calculate sensitivity and specificity of predicting progression to moderate fibrosis for the predictive risk scores at 5 years including a bootstrap to calculate confidence intervals (CIs). All statistical analyses were conducted using SAS version 9.4 software (SAS Institute, Cary, North Carolina).

RESULTS

At baseline the median age was 49 years (IQR, 44–53 years), 72% were male, and 87% were African American (Table 1). Half (52%) were actively injecting and 14% had an AUDIT score indicating alcohol dependence. The median duration of drug injection was 27.5 years (IQR, 20–35 years). Thirty-five percent were HIV coinfected, 52% of whom were not on antiretroviral therapy. Nearly two-thirds (n = 604 [63%]) had no or mild/fibrosis, 218 (23%) had moderate fibrosis, and 142 (15%) had severe fibrosis/cirrhosis; 4%, 14%, and 21% of those with no/mild, moderate, and severe fibrosis, respectively, had diabetes.

Table 1.

Characteristics of Study Sample at First Liver Stiffness Measurement (N = 964)

Characteristic	No. (%) or Median (IQR)
Age, y	49 (44–53)
Sex
Male	690 (71.6)
Female	274 (28.4)
Race
White and other race	126 (13.1)
Black	838 (86.9)
Drug use in past 6 months
No drug use	362 (37.6)
Noninjection drug use only	103 (10.7)
Only injection drug use only	153 (15.9)
Both noninjection and injection drugs	345 (35.8)
Years since first injection drug use	27.5 (20–35)
Alcohol usea
No/mild alcohol use	711 (73.8)
Hazardous alcohol use	118 (12.2)
Dependence	135 (14.0)
HIV status
Negative	627 (65.0)
Positive	337 (35.0)
HIV RNA level (among HIV positive)
<50 copies/mL (undetectable)	122 (36.5)
50–9999 copies/mL	93 (27.8)
≥10000 copies/mL	119 (35.6)
CD4 count (among HIV positives)
≥500 cells/µL	63 (18.8)
200–499 cells/µL	156 (46.6)
<200 cells/µL	116 (34.6)
CD4 nadir (among HIV positives)
≥500 cells/µL	25 (7.5)
200–499 cells/µL	118 (35.2)
<200 cells/µL	192 (57.3)
HAART usea (among HIV positives)
Yes	140 (48.3)
No	150 (51.7)
HCV RNA level
<6 log10 IU/mL	343 (35.6)
≥6 log10 IU/mL	621 (64.4)
Body mass index, kg/m2
<25	489 (52.4)
25–30	296 (31.7)
>30	148 (15.9)
Hypertensionb
No	492 (51.0)
Yes	472 (49.0)
Diabetesc
No	876 (91.1)
Yes	86 (8.9)
Baseline liver stiffness measurement, kPa
<8	604 (62.7)
8–12.3	218 (22.6)
>12.3	142 (14.7)
FIB-4 scored
<1.45	415 (47.7)
1.45–3.25	371 (42.6)
>3.25	85 (9.8)

Abbreviations: FIB-4, Fibrosis-4; HAART, highly active antiretroviral therapy; HCV, hepatitis C virus; HIV, human immunodeficiency virus; IQR, interquartile range.

^aDetermined by Alcohol Use Disorders Identification Test.

^bSelf-reported of taking high blood pressure medication prescribed by health care provider in the last 6 months or if systolic blood pressure reading >150 mm Hg or diastolic blood pressure >90 mm Hg at baseline liver stiffness measurement (LSM) visit.

^cSelf-reported diabetic medication use in the last 6 months or if A1C level >6.5% within a year of baseline LSM visit.

^dClosest available to baseline LSM (n = 871).

Mortality

Over a median of 5.9 years (IQR, 4.3–6.5) after the first LSM, 155 deaths were observed (mortality rate, 3.06 deaths per 100 person-years [PY]). All-cause mortality was highest among participants whose first LSM was consistent with severe fibrosis/cirrhosis (6.21 deaths/100 PY). Mortality was also significantly elevated among participants whose first LSM was consistent with moderate fibrosis (3.59 deaths/100 PY) compared to those whose first LSM was <8.0 kPa (2.21 deaths/100 PY; P = .014). When LSMs were grouped into smaller intervals, similar trends in increasing mortality risk were observed (P_trend <.001) (Figure 2).

Figure 2.

All-cause mortality rates (A) and nonaccidental mortality rates (B) according to baseline liver stiffness measurement. Abbreviations: CI, confidence interval; PY, person-years.

In multivariable models adjusting for only sociodemographics (model 1) and substance use, (model 2) mortality risk remained significantly elevated among individuals with moderate fibrosis and severe fibrosis/cirrhosis (Table 2). After adjusting for HIV RNA load, BMI, and diabetes and hypertension comorbidities, mortality risk remained significantly elevated among those with severe fibrosis/cirrhosis (adjusted hazard ratio [aHR], 2.19 [95% CI, 1.44–3.33]); however, the association with moderate fibrosis was attenuated and did not retain statistical significance (model 4), (aHR, 1.39 [95% CI, .92–2.09]). Compared to all-cause mortality, the adjusted measure of association between fibrosis/cirrhosis and nonaccidental mortality was stronger among those with severe fibrosis/cirrhosis (aHR, 2.75 [95% CI, 1.72–4.40]) and moderate fibrosis (aHR, 1.61 [95% CI, 1.01–2.57]) after adjusting for sociodemographic, behavioral, and comorbid factors (model 4, Table 3). Similar trends were observed when using an alternative, validated method to ascertain liver fibrosis severity. Compared to individuals with FIB-4 <1.45, those with FIB-4 >3.25 and between 1.45 and 3.25 had a 4.44-fold (95% CI, 2.52–7.82; P < .0001) and 1.51-fold increased risk of mortality (95% CI, .89–2.56; P = .124), respectively, in an unadjusted model.

Table 2.

Factors Associated With All-Cause Mortality (N = 964 Participants, 155 Deaths)

Covariate	Model 1 aHR (95% CI)	P Value	Model 2 aHR (95% CI)	P Value	Model 3 aHR (95% CI)	P Value	Model 4 aHR (95% CI)	P Value
LSM, kPa
<8.0	1.00		1.00		1.00		1.00
8.0–12.3	1.51 (1.02–2.23)	.039	1.55 (1.05–2.29)	.028	1.42 (.96–2.11)	.079	1.39 (.92–2.09)	.12
>12.3	2.47 (1.68–3.63)	<.001	2.55 (1.72–3.77)	<.001	2.21 (1.49–3.29)	<.001	2.19 (1.44–3.33)	<.001
Age per 10-y increase	1.51 (1.19–1.92)	<.001	1.59 (1.24–2.03)	<.001	1.85 (1.43–2.39)	<.001	1.64 (1.24–2.16)	<.001
Sex
Male	1.00		1.00		1.00		1.00
Female	1.36 (.97–1.92)	.078	1.42 (1.00–2.02)	.049	1.35 (.95–1.92)	.095	1.33 (.92–1.92)	.13
Race
White and other race	1.00	.71	1.00	.78	1.00	.27	1.00
Black	0.90 (.50–1.60)		0.92 (.51–1.65)		0.72 (.40–1.29)		0.73 (.40–1.34)	.31
Drug use in past 6 mo
No drug use			1.00		1.00		1.00
Noninjection drug use only			0.93 (.53–1.63)	.79	0.97 (.55–1.71)	.92	0.89 (.51–1 .58)	.70
Only injection drug use			1.22 (.75–1.99)	.42	1.20 (.74–1.97)	.46	1.00 (.60–1.67)	.99
Both noninjection and injection drugs			1.28 (.86–1.92)	.22	1.29 (.86–1.93)	.23	1.04 (.68–1.58)	.87
Alcohol usea
No/mild alcohol use			1.00		1.00		1.00
Hazardous alcohol			1.23 (.75–2.03)	.42	1.14 (.69–1.89)	.60	1.15 (.70–1.92)	.58
Dependence			1.29 (.83–2.01)	.26	1.25 (.80–1.95)	.32	1.27 (.81–2.01)	.30
HIV RNA level
Negative					1.00		1.00
<50 copies/mL					1.55 (.94–2.55)	.087	1.58 (.95–2.63)	.077
50–9999 copies/mL					2.23 (1.37–3.63)	.001	1.96 (1.18–3.25)	.009
≥10000 copies/mL					3.55 (2.35–5.35)	<.001	3.28 (2.15–5.02)	<.001
Body mass index, kg/m2
<25							1.00
25–30							0.56 (.38–.83)	.004
>30							0.45 (.26–.79)	.006
Diabetes
No							1.00
Yes							1.65 (1.01–2.71)	.046
Hypertension
No							1.00
Yes							1.46 (1.02–2.10)	.039

Model 1, sociodemographics; model 2, sociodemographics and behavioral; model 3, sociodemographics, behavioral, and HIV infection; model 4, sociodemographics, behavioral, HIV infection, and comorbidities.

Abbreviations: aHR, adjusted hazard ratio; CI, confidence interval; HIV, human immunodeficiency virus; LSM, liver stiffness measurement.

^aDetermined by Alcohol Use Disorders Identification Test.

Table 3.

Factors Associated With Nonaccidental Mortality (N = 964 Participants, 121 Deaths)

Covariate	Model 1 aHR (95% CI)	P Value	Model 2 aHR (95% CI)	P Value	Model 3 aHR (95% CI)	P Value	Model 4 aHR (95% CI)	P Value
LSM, kPa
<8.0	1.00		1.00		1.00		1.00
8.0–12.3	1.80 (1.15–2.80)	.001	1.81 (1.16–2.83)	.009	1.66 (1.06–2.59)	.027	1.61 (1.01–2.57)	.045
>12.3	3.15 (2.05–4.85)	<.001	3.18 (2.05–4.94)	<.001	2.69 (1.73–4.20)	<.001	2.75 (1.72–4.40)	<.001
Age per 10-y increase	1.49 (1.14–1.96)	.004	1.52 (1.15–2.02)	.004	1.83 (1.36–2.47)	<.001	1.64 (1.19–2.26)	.003
Sex
Male	1.00		1.00		1.00		1.00
Female	1.63 (1.11–2.37)	.012	1.67 (1.14–2.46)	.009	1.62 (1.10–2.40)	.015	1.60 (1.06–2.42)	.024
Race
White and other race	1.00		1.00		1.00		1.00
Black	1.28 (.61–2.70)	.52	1.28 (.60–2.71)	.52	0.97 (.46–2.06)	.93	1.07 (.48–2.40)	.87
Drug use in past 6 mo
No drug use			1.00		1.00		1.00
Noninjection drug use only			0.80 (.43–1.52)	.50	0.82 (.43–1.56)	.55	0.72 (.38–1.37)	.32
Only injection drug use			1.15 (.71–1.97)	.61	1.11 (.64–1.91)	.71	0.86 (.49–1.52)	.61
Both noninjection and injection drugs			1.02 (.67–1.61)	.94	1.02 (.64–1.62)	.94	0.77 (.48–1.24)	.28
Alcohol usea
No/mild alcohol use			1.00		1.00		1.00
Hazardous alcohol use			1.09 (.60–1.97)	.78	0.99 (.55–1.80)	.97	0.99 (.54–1.81)	.97
Dependence			1.07 (.62–1.82)	.82	1.01 (.59–1.73)	.97	1.00 (.57–1.75)	.99
HIV RNA level
Negative					1.00		1.00
<50 copies/mL					1.53 (.87–2.68)	.140	1.54 (.87–2.74)	.139
50–9999 copies/mL					2.75 (1.62–4.69)	<.001	2.37 (1.37–4.11)	.002
≥10000 copies/mL					3.94 (2.47–6.27)	<.001	3.61 (2.23–5.86)	<.001
Body mass index, kg/m2
<25							1.00
25–30							0.43 (.27–.69)	<.001
>30							0.37 (.20–.69)	.002
Diabetes
No							1.00
Yes							2.05 (1.21–3.45)	.007
Hypertension
No							1.00
Yes							1.31 (.87–1.97)	.194

Model 1, sociodemographics; model 2, sociodemographics and behavioral; model 3: sociodemographics, behavioral, and HIV infection; model 4: sociodemographics, behavioral, HIV infection, and comorbidities.

Abbreviations: aHR, adjusted hazard ratio; CI, confidence interval; HIV, human immunodeficiency virus; LSM, liver stiffness measurement.

^aDetermined by Alcohol Use Disorders Identification Test.

Progression to Moderate Fibrosis

As mortality was elevated even in persons with moderate fibrosis, we evaluated whether this transition could be predicted with sufficiently high diagnostic accuracy to justify the practice of some insurers who deny treatment to this group. We studied 2719 LSMs over a median of 5.7 years (IQR, 3–7 years) per person. Of 331 individuals with no/mild fibrosis at baseline, 63 (19%) had evidence of significant progression measured by ≥2 subsequent LSMs ≥8.0 kPa (incidence rate, 3.55/100 PY). In univariable analysis, high HIV RNA (≥10000 copies/mL), and higher initial LSM at T_-3 were associated with an increased risk of progressing to ≥8.0 kPa (HR, 2.31 [95% CI, 1.32–4.05] per kPa and 1.30 per kPa [95% CI, 1.10–1.55], respectively). In multivariable analysis (Table 4), which did not include the visit-to-visit changes (model 1), high HIV RNA level (≥10000 copies/mL) (aHR, 2.75 [95% CI, 1.28–5.92]; P = .001), and being overweight (aHR vs normal BMI, 1.88 [95% CI, 1.09–3.26]) were significantly associated with an increased risk of progression, but had low predictive accuracy (C-statistic = 0.66 [95% CI, .60–.73]). The sensitivity for predicting transition from no/mild to moderate fibrosis at 5 years among those whose risk score was in the top quintile was 32% and the specificity was 83%.

Table 4.

Factors Associated With Progressing From No/Mild Fibrosis (<8.0 kPa) to Moderate Fibrosis (≥8.0 kPa)^a

Covariate	Model 1 aHR (95% CI)	P Value	Model 2 aHR (95% CI)	P Value	Model 3 aHR (95% CI)	P Value	Model 4 aHR (95% CI)	P Value
Age per 10-y increase	1.05 (.68–1.62)	.84	1.06 (.68–1.65)	.79	1.04 (.67–1.63)	.87	1.07 (.69–1.67)	.76
Sex
Male	1.00		1.00		1.00		1.00
Female	0.88 (.47–1.65)	.69	0.89 (.48–1.68)	.72	0.92 (.49–1.72)	.79	0.88 (.47–1.66)	.69
Race
White and other race	1.00		1.00		1.00		1.00
Black	1.94 (.44–8.52)	.38	1.72 (.39–7.65)	.48	1.67 (.37–7.44)	.50	1.63 (.37–7.25)	.52
Drug use
No drug use	1.00		1.00		1.00		1.00
Noninjection drug use only	0.84 (.38–1.89)	.68	0.86 (.39–1.91)	.71	0.86 (.39–1.92)	.72	0.88 (.40–1.97)	.76
Only injection drug use	1.33 (.65–2.71)	.44	1.31 (.64–2.69)	.46	1.27 (.62–2.60)	.52	1.32 (.65–2.70)	.44
Noninjection and injection drugs	1.06 (.51–2.23)	.87	1.06 (.51–2.23)	.87	1.09 (.52–2.28)	.83	1.05 (.50–2.21)	.89
AUDIT
No/mild hazardous alcohol use	1.00		1.00		1.00		1.00
Hazardous alcohol use	1.53 (.72–3.26)	.27	1.59 (.75–3.37)	.22	1.62 (.76–3.44)	.21	1.58 (.75–3.36)	.23
Dependence	1.11 (.47–2.58)	.82	1.02 (.44–2.38)	.97	1.00 (.43–2.34)	.99	0.96 (.41–2.26)	.93
HIV RNA level
Negative	1.00		1.00		1.00		1.00
<50 copies/mL	0.73 (.29–1.88)	.52	0.72 (.28–1.86)	.50	0.72 (.28–1.85)	.50	0.73 (.29–1.89)	.52
50–9999 copies/mL	2.10 (.92–4.81)	.078	2.20 (.96–5.05)	.062	2.15 (.94–4.95)	.071	2.18 (.95–5.00)	.066
≥10000 copies/mL	2.75 (1.28–5.92)	.001	2.58 (1.18–5.61)	.017	2.42 (1.10–5.33)	.029	2.53 (1.15–5.55)	.021
Body mass index, kg/m2
<25	1.00		1.00		1.00		1.00
25–30	1.88 (1.09–3.26)	.024	1.91 (1.01–3.32)	.021	1.94 (1.12–3.38)	.019	1.90 (1.10–3.29)	.022
>30	0.95 (.35–2.58)	.92	0.89 (.33–2.42)	.81	0.79 (.29–2.16)	.64	0.85 (.31–2.32)	.75
LSM at T-3b			1.28 (1.07–1.54)	.008	1.57 (1.24–2.00)	<.001	1.45 (1.14–1.84)	.002
∆T-1, T-2c			1.32 (.42–4.12)	.63
∆T-1, T-3d					41.33 (3.15–542.31)	.005
∆T-2, T-3e							2.87 (.84–9.75)	.091

Model 1, all covariates except LSM changes; model 2: all covariates, LSM, LSM changes between T_-1 and T_-2; model 3, all covariates, LSM, LSM changes between T_-1 and T_-3; model 4, all covariates, LSM, LSM changes between T_-2 and T_-3.

Abbreviations: aHR, adjusted hazard ratio; AUDIT, Alcohol Use Disorder Identification Test; CI, confidence interval; HIV, human immunodeficiency virus; LSM, liver stiffness measurement.

^aAt least 2 visits where LSM >8.0 kPa; 331 individuals, 2719 LSMs.^bLSM at 3 visits before event.

^cSix-month change, 1 visit (approximately 6 months) before the event.

^dOne-year change, 1 visit (approximately 6 months) before the event.

^eSix-month change, 2 visits (approximately 12 months) before the event.

In models that included combinations of LSM visit-to-visit changes, only a 1-year change in LSM was significantly associated with progression. For every 1-kPa increase in the 1-year change in LSM, the risk of progressing to moderate fibrosis increased >40-fold (aHR, 41.3 [95% CI, 3.15–542.3]), after adjusting for sociodemographics, substance use behaviors, BMI, and LSM at T_-3. However, this change had only fair predictive accuracy (C-statistic: 0.72 [95% CI, .66–.78]). Associations with 6-month changes were not significantly associated with progression regardless of whether the 6-month change occurred 1 visit (∆T_-1, T_-2: aHR, 1.32 [95% CI, .42–4.12]) or 2 visits (∆T_-2, T_-3: aHR, 2.87 [95% CI, .84–9.75]) prior to the progression event. The C-statistics for the 6-month changes were similarly in the fair prognostic accuracy range (∆T_-1, T_-2 C-statistic: 0.70 [95% CI, .64–.76]; ∆T_-2, T_-3 C-statistic: 0.71 [95% CI, .65–.76]).

DISCUSSION

In this large sample of PWID with chronic HCV infection, we observed substantial mortality. Although increased mortality was evident among individuals with severe fibrosis/cirrhosis, we observed some increased risk of mortality even among those with moderate fibrosis. These findings, and our inability to identify with sufficiently high prognostic accuracy individuals who would transition from a lower mortality risk state (minimal liver disease) to a higher mortality risk state (moderate or severe liver disease), may not support withholding HCV treatment until that transition occurs.

Prior studies have demonstrated an association between severe fibrosis/cirrhosis and mortality [18, 19]. Indeed, surveillance data released by the CDC indicate that in 2014, HCV infection killed an estimated 20000 persons in the United States [20]. Much of the mortality is due to liver failure and hepatocellular carcinoma. But, as we observed, persons with HCV infection also have increased mortality that may be unrelated to HCV-induced liver disease and may be the result of a negative effect of HCV on other organ systems. Indeed, excess mortality due to circulatory and renal diseases was reported in a cohort of people chronically infected with HCV [21]. Here we report a higher rate of mortality in persons with moderate liver disease, compared to no/mild fibrosis which may be attributed to chronic inflammation resulting from chronic HCV infection and comorbidities [7]. Furthermore, the increased mortality among those with moderate fibrosis at baseline may in part be due to progression to cirrhosis in this group. Although there were too few deaths to evaluate this hypothesis, higher mortality was observed for those who progressed from no/mild fibrosis to moderate fibrosis (8% of whom then progressed to cirrhosis) vs those who remained in the no/mild fibrosis state during the follow-up period (1.33 deaths per 100 PY vs 1.06 deaths per 100 PY).

While the overall association between baseline moderate fibrosis and mortality attenuated slightly after including BMI, diabetes, and hypertension in the model for nonaccidental deaths, it is possible that these comorbidities, particularly diabetes, are causally associated with fibrosis and mortality [22]. Much of the mortality observed was not liver related, which reflects both the nature of the population (all had a history of drug use) and the challenges in ascertaining cause of death from death certificates.

Consistent with previous studies, we found that high HIV RNA level was associated with an increased risk of progression from no/low to moderate fibrosis [8, 23, 24]. High HIV RNA level is a surrogate for untreated HIV infection, and use of antiretroviral therapy has been associated with attenuated liver fibrosis progression in HIV-infected persons [8, 25]. However, while this association was statistically significant, high HIV RNA level did not discriminate among persons who did and did not progress with sufficiently high accuracy. Moreover, no other demographic, behavioral, or clinical factor, nor changes in LSMs themselves had high predictive accuracy for fibrosis progression.

Successful treatment of HCV is associated with 80% reduction in mortality, as well as reduced incidences of liver failure and hepatocellular carcinoma [26]. This association has been demonstrated in several settings, including among individuals with moderate liver disease [27, 28]. Although we were not able to evaluate the impact of sustained virologic response in our population, our data in combination with these other findings support the recommendation of AASLD/IDSA to treat nearly all HCV-infected persons immediately, a strategy also found to be cost-effective [29].

Withholding medical treatment based on disease stage implies that there is a “safe” disease stage. Additionally, it is assumed that the “safe” stage and transitions out of that stage can be accurately detected. Whether detected by liver biopsy, LSM, or blood markers, liver fibrosis estimates cannot differentiate mild from moderate fibrosis with sensitivity >80% [30, 31]. Likewise, progression of liver fibrosis was not predicted with sufficiently high diagnostic accuracy in most other studies [23, 32]. These scientific and clinical realities should be carefully considered when formulating evidence-based approaches to delivering HCV treatment.

The experience of the ALIVE cohort may differ from persons with chronic HCV in other settings, including those outside the United States. For example, this cohort had a higher prevalence of HIV and HCV than national CDC estimates [33] and findings from the National Health and Nutrition Examination Survey [2]. Additionally, more persons in this setting continued to use illicit drugs than in many primary care or specialty clinical practices. Nonetheless, >80% of HCV infections in the United States are due to illicit drug use and an estimated 75–80% of those infected were born between 1945–1965, similar to ALIVE cohort members. Moreover, while there was high competing mortality, other unrelated causes of death should reduce our ability to detect an association based on liver disease stage rather than create a false finding. Indeed, elimination of accidental deaths strengthened the estimated mortality association. While we used death certificate data to determine cause of death, these data are often not comprehensive. For example, we considered deaths due to “chronic drug use” to be accidental because physicians may have ascribed this cause of death when they could not determine a precise natural cause of death. However it is possible that deaths due to chronic drug use may have been due to underlying liver disease, which has been shown to be underestimated in death certificates [34]. Some members of the ALIVE cohort have received successful HCV treatment and would have been excluded by having undetectable HCV RNA. However, the course of person-time evaluated was prior to the availability of all-oral medications in 2015. During that interval, <5% were cured by treatment [35]. This analysis was based chiefly on elastography. However, we observed high concordance between our LSM values and blood tests such as FIB-4. As these tests are often used in clinical practice, these data would still be germane to the subset of Medicaid-supported patients currently denied HCV treatment in many states that require a minimum stage of F3 or F4 to obtain approval for treatment [5].

In summary, we observed increased mortality among persons with severe and moderate fibrosis, and transitions to moderate fibrosis could not be predicted with high accuracy. These data support the AASLD/IDSA guidelines for treatment of all persons with chronic HCV infection and do not support withholding treatment from those with mild disease.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. Conception and design: J. A. C., D. L. T., and S. H. M. Analysis and interpretation of data: J. A. C., D. L. T., J. A., S. H. M. Drafting of the article: J. A. C., J. A., S. H. M. Final approval of the article: J. A. C., D. L. T., J. A., M. S. S., G. D. K., S. H. M. Provision of study materials or patients: G. D. K. and S. H. M. Statistical expertise: J. A. C., J. A., S. H. M.

Acknowledgments. We thank all of the study participants.

Disclaimer. The funding agencies had no role in the design, analysis, or interpretation of these findings.

Financial support. This work was supported by the National Institute on Drug Abuse (grant numbers R01DA012568, R37DA013806, and K24DA034621) and the National Institute of Allergy and Infectious Diseases (grant numbers P30AI094189 and T32AI102623).

Potential conflicts of interest. M. S. S. has received research grants from AbbVie, Gilead, Janssen, and Merck, fees for consultancy from AbbVie, BMS, Cocrystal, Gilead, Janssen, Merck, and Trek, and payment for development of educational presentations from ViralEd, Practice Point, and Clinical Care Options. All other authors declare no conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.